Remaining unspecified components are set to 0, 0, 1, according to spec.
Also cleaned up and simplified the internals, added debug output
operators for attribute component count and types and tested the whole
thing.
The enum is finally completely reimplemented for specific cases with
proper values (Mesh attribute types, index types, pixel types, ...). The
rest currently remaining in TypeTraits.h is used only in Mesh and
AbstractShaderProgram, probably will be moved there.
(De)normalization from/to [0, 1] for unsigned types and from/to [-1, 1]
for signed types. -1.0 is always denormalized to min+1.
Also updated some tests for Color, as they depended on previous
imprecise implementation.
Removing of another <*stream> #include leads to more compilation time
saving, now from ~5:12 to ~4:55. Another compilation time improvements
will now be possible only by using Clang's modules, I don't know where
to optimize further (except for getting rid of <sstream> in tests).
Optimalizations in Corrade::TestSuite and Corrade::Utility::Debug leaded
to significant reduction of compilation time - on my machine it was
~5:38 before with building of unit tests enabled, now only ~5:00.
Before it has to be done with overly verbose and cumbersome Java-style:
Vector4 vec;
vec.setX(vec.x() + 5);
vec = Vector4(vec.xyz()*2, vec.w());
Now it can be done this way:
vec.x() += 5;
vec.xyz() *= 2;
They aren't much useful in Math namespace and here it allows to handle
also Color3/Color4 types. Moreover all RGBA component names were removed
from Math namespace, so it would be inconsistent to have them there only
in swizzle() functions.
They can be used for both floating-point (normalized) and integral
(denormalized) representation and support conversion between them.
HSV conversion is done always using floating-point types.
Getters for RGB(A) were removed from Vector3 and Vector4, they are now
part of Color3 and Color4. Because of this, Framebuffer now accepts
Color instead of Vector.
Now we can use CORRADE_COMPARE() almost exclusively for comparing
Vector, Matrix instances and all containers from STL. Tests running time
for all 31 tests is now around 0.1 sec instead of two seconds like
before.
Disabled compilation of all benchmarks, because TestSuite cannot handle
them yet.
PKGBUILD from AUR now can shamelessly use check() function without
unnecessary dependencies.